KR20150087327A - Pattern shape evaluation method, semiconductor device manufacturing method, and pattern shape evaluation device - Google Patents

Abstract

The cross-sectional shape or the three-dimensional shape of the circuit pattern is estimated and evaluated only from the plane of the circuit pattern observed from above the wafer. The present invention relates to a method for manufacturing a semiconductor device, which irradiates a convergent energy line from a direction substantially perpendicular to a main surface of a substrate on which a three-dimensional structure is formed on an upper surface, scans an upper surface of the substrate, A process of detecting and / or measuring the intensity of an energy ray reflected or scattered by the structure and obtaining an image of the top view of the structure; A process of obtaining uncertainty information on the scattering intensity due to the concavo-convex shape of the surface of the structure, a process of obtaining the inclination angle? Of the surface of the structure on the basis of the obtained uncertainty information, .

Description

TECHNICAL FIELD [0001] The present invention relates to a pattern shape evaluation method, a semiconductor device manufacturing method, and a pattern shape evaluation apparatus. BACKGROUND OF THE INVENTION 1. Field of the Invention [0002]

The present invention relates to, for example, a detailed shape evaluation method by non-destructive observation and image processing using a scanning microscope, an apparatus therefor, and a method of manufacturing a semiconductor device employing the technique.

2. Description of the Related Art Semiconductor integrated circuits (LSIs) are becoming more sophisticated and highly integrated as circuit patterns become finer. At present, the line width of the minimum circuit pattern of the LSI endmost is 30 nm (nanometer) or less. In order to secure the performance of the LSI, it is necessary to strictly set these circuit dimensions (for example, With a precision of less than 10% of the design value). Currently, a scanning electron microscope (SEM) is widely used to measure the circuit dimensions. Non-Patent Document 1 describes a field emission type electron microscope (CD-SEM: Critical Dimension SEM) for circuit dimension measurement for observing a wafer from above.

The CD-SEM is used for measuring various characteristic quantities in addition to line width measurement of a semiconductor circuit. For example, it is known that the edge of the circuit pattern has unevenness called line edge roughness (LER), which adversely affects circuit performance. The CD-SEM is widely used for the measurement of the LER. For example, Patent Literature 1 describes a measurement method thereof.

On the other hand, there is a demand for knowing the three-dimensional shape of the element structure formed by lamination and patterning on the wafer. Particularly, in the LSI mass production process, it is desirable to do this non-destructively without observing the cross section. However, AFM or optical method (Scatterometry) is generally used as a method for this.

The AFM is a method of measuring the shape of the concavo-convexity of the surface of a sample by scanning the probe with a fine tip so that the interatomic force between the tip of the probe and the surface of the sample becomes constant. The details thereof are described, for example, in Patent Document 2.

Scatterometry is a method of measuring the cross-sectional shape of a three-dimensional structure by measuring the wavelength or diffraction angle dependence of the reflection diffracted light by inputting light into a pattern having a periodic three-dimensional structure and comparing it with diffraction angle dependence obtained by calculation for various cross- . Scatterometry is described in Non-Patent Document 2, for example.

Similar to Scatterometry, there is a model-based library (MBL) method to estimate the cross-sectional shape using SEM. In the MBL method, by comparing the secondary electron detection signal intensity distribution obtained by scanning a converged electron beam to a sample and the secondary electron signal intensity distribution obtained by calculation for various sectional shapes in advance, Estimate the cross-sectional shape. The MBL is described in, for example, Patent Document 3 or Non-Patent Document 3.

As a method of measuring the three-dimensional structure using SEM, there is tilt-SEM. This method assumes a three-dimensional shape based on the principle of a stereoscopic image from a plurality of images obtained by incidence of an electron beam from different angles with respect to a wafer. The tilt-SEM is described in, for example, Patent Document 4.

Japanese Patent Application Laid-Open No. 2006-215020 Japanese Patent Application Laid-Open No. 2009-257937 Japanese Patent Application Laid-Open No. 2007-227618 Japanese Patent Application Laid-Open No. 2005-183369

Hitachi Review vol. 60 (2011), No. 5 pp. 203-209 Solid State Technology, Vol. 54, Issue 8, (2011) Proceedings of SPIE Vol. 5375 (SPIE, Bellingham, WA, 2004) Dimensional Metrology of Resist Lines using a SEM Model-Based Library Approach

In the process of manufacturing a semiconductor integrated circuit (LSI), it is important to manage a circuit pattern or a cross-sectional shape such as a resist pattern for forming the circuit pattern, or a three-dimensional shape within a predetermined design range. It is necessary to set the manufacturing conditions such that these shapes conform to the design values or, when the shape deteriorates, to quickly detect these shapes and feed back to the manufacturing process to suppress quality deterioration.

However, up to now, there has been a technical problem that there is no method for easily measuring the sectional shape or three-dimensional shape of a fine circuit pattern of arbitrary shape with non-destructive, high accuracy. For example, the CD-SEM can measure the dimensions of fine circuit patterns of arbitrary shapes in a non-destructive, high-precision, and simple manner. However, since the planar shape of the circuit is observed from above the wafer, I have an assignment.

In general, therefore, a method is used in which a wafer is broken to expose an end face of a circuit pattern and the cross-sectional shape is observed with an electron microscope. However, it is difficult to apply this method to product wafers. Thus, conventionally, it has been considered difficult to estimate the cross-sectional shape using the CD-SEM, and various methods as described above have been attempted.

However, in any of these methods, there are technical problems as follows.

For example, in the method using AFM, there is a problem that it becomes difficult to measure the shape because the size of the circuit pattern is reduced, the probe can not enter between the patterns.

Further, in the method using scatterometry or MBL, it is necessary to prepare the library as a library by previously calculating the expected measurement results for various sectional shapes. However, this requires a large amount of calculation, and there is a problem that the applicable shape is limited to a relatively simple shape such as a trapezoid. Particularly, in the method based on scatterometry, since the measurement pattern is limited to a periodic pattern that is constantly present within a wide region of several tens of microns (angles), it is difficult to estimate the cross-sectional shape of an irregular pattern such as a circuit of a logic LSI It is difficult. Further, this method has a problem that it is necessary to secure a wide area dedicated for the measurement pattern.

In addition, in the method using tilt-SEM, a special electro-optical system for changing the angle of incidence of the convergent electron beam is required, and in general, the performance such as resolution deteriorates. Further, there is also a problem that the apparatus is made larger in order to tilt the stage, a problem that it takes time to measure, and the like.

Therefore, the present invention provides a technique capable of estimating the cross-sectional shape of an arbitrary structure formed on the upper surface of the substrate to nondestructive and highly accurate, while using only the observation image from the upper surface of the substrate obtained using the charged particle beam device .

In order to solve the above problems, for example, the configuration described in claims is adopted. (A) irradiating a convergent energy beam from a direction substantially perpendicular to a main surface of a substrate on which a three-dimensional structure is formed on an upper surface, scanning the upper surface of the substrate, and A process of detecting and / or measuring the intensity of an energy ray reflected or scattered by the secondary energy ray or the substrate and the structure and obtaining an image of the top view of the structure, and (b) (C) obtaining uncertainty information of scattering intensity due to the concavo-convex shape of the surface of the structure from the energy ray irradiation position and the measured intensity; and (c) calculating an inclination angle? Of the surface of the structure based on the obtained uncertainty information. And (d) estimating the three-dimensional shape of the structure based on the obtained inclination angle?.

According to the present invention, it is possible to estimate and evaluate the sectional shape or the three-dimensional shape of the structure formed on the upper surface of the substrate only from the upper surface observation side of the substrate. The problems, constitutions and effects other than those described above are revealed by the following description of the embodiments.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a conceptual diagram illustrating an electron microscope observation process. FIG.
2 is a conceptual view showing a feature of an observed structure.
3 is a conceptual diagram for explaining the principle of the present invention;
4 is a schematic view showing the characteristics of a cross-sectional shape of a sample for explaining the first embodiment;
Fig. 5 is a characteristic diagram schematically showing measurement results of each sample shown in Fig. 4; Fig.
Fig. 6 is another characteristic diagram schematically showing measurement results for each sample shown in Fig. 4; Fig.
7 is a characteristic diagram showing the influence of image noise.
8 is a characteristic diagram showing an analysis result of the measurement result shown in Fig.
9 is a principle view showing the relationship between the inclination angle of the structure surface and the fluctuation width of the edge detection point.
10 is a principle view for explaining another model for obtaining the relationship between the inclination angle of the structure surface and the fluctuation width of the edge detection point.
11 is a schematic diagram showing the influence of the variation of the side surface of the structure on the detection signal intensity for an electron beam incident on a point relatively far from the side.
12 is another characteristic diagram showing an analysis result of the measurement result shown in Fig.
Fig. 13 is a characteristic diagram showing the result of estimation of the cross-sectional shape of each sample shown in Fig. 4; Fig.
FIG. 14 is a flowchart for explaining a processing procedure in the first embodiment. FIG.
15 is a schematic diagram showing a configuration example of an apparatus used in the first embodiment;
16 is a schematic diagram of a three-dimensional shape estimation method according to the second embodiment;
17 is a schematic diagram of a process monitor according to the third embodiment.
18 is another conceptual diagram showing a feature example of the observed structure.
FIG. 19 is a flowchart for explaining a processing procedure in the second embodiment. FIG.
20 is a characteristic diagram for explaining analysis results according to the second embodiment;
FIG. 21 is a conceptual diagram for explaining the fifth embodiment. FIG.
22 is a conceptual diagram for explaining a fifth embodiment;

Hereinafter, embodiments of the present invention will be described with reference to the drawings. The embodiment of the present invention is not limited to the following embodiment, but various modifications are possible within the scope of the technical idea. For example, in the following description, a case of observing a sample using an electron microscope is described. However, the case of observing a sample using a charged particle beam apparatus other than a focusing (focused) ion beam apparatus, can do.

(Edge Variation Factors and Component Decomposition on Three-Dimensional Shape Observation)

First, a process of forming an observation image by an electron microscope used in the present invention will be described with reference to Fig. For the sake of simplicity, a substantially cubic structure (for example, a semiconductor or a resist pattern) formed on the substrate surface is considered. Let the substrate surface be the x-y plane, and the direction in which the edge of the structure extends (the edge direction) is the y direction. Irradiating the substrate with an electron beam converged sufficiently narrower than the characteristic dimension of the structure on the substrate in a direction (z direction) substantially perpendicular to the substrate surface, and scanning the substrate on the substrate in a direction (x direction) substantially perpendicular to the edge direction do. The electrons incident on the substrate or the structure are scattered in the substrate or the structure to emit secondary electrons or are directly reflected (or backscattered) and emit a part thereof to the outside of the substrate or the structure. The emission amount of the secondary electron or the reflection electron (hereinafter referred to as " secondary electron or the like ") increases when the electron beam enters the convex portion (or the upper portion of the corner) of the convex shape.

Thus, when the secondary electrons or the like emitted during the scanning of the electron beam are detected by the detector and the detected intensity is plotted against the irradiation position x of the electron beam in the scanning direction, the secondary electron And the like are obtained. Generally, when the detection signal intensity distribution is normalized to the maximum value of the detection signal and the position where the detection signal intensity distribution is cut at a predetermined threshold level or the position where the inclination of the detection signal intensity distribution becomes maximum, . The pattern dimension is measured from the distance between the two pattern boundary positions. Further, the scanning is performed at a position different from the edge direction (y direction), and the edge shape of the pattern is obtained by connecting the obtained pattern / non-pattern boundary. Usually, the edge shape thus obtained exhibits a concavo-convex shape along the edge direction. The size of this irregularity is called line edge roughness. The inventors of the present invention have studied the cause of the roughness, and have found that from the detected signal intensity distribution in a plane (hereinafter referred to as " substrate surface ") including a direction parallel to the pattern edge and a direction perpendicular to the pattern edge, We have devised a method for estimating the three-dimensional shape of a pattern in a direction.

The above-described line edge roughness is generally considered to be due to the deviation of the edge position of the structure as shown in the top view (a) of Fig. The actual structure is not a cube, and its side wall is inclined as shown in the upper view (a) of Fig. 2, or has irregularities (surface roughness) on the surface as shown in the interrupted view (b) of Fig. It is considered that the detected signal intensity distribution is affected by the change of the inclination angle and the unevenness of the surface.

When the secondary electron detection signal intensity is measured by scanning an electron beam in the x direction from a starting point defined at a constant sampling interval with respect to the y direction, the secondary electron detection signal intensity of 2 Dimensional distribution is obtained. Here, the two-dimensional distribution of the secondary electron detection signal intensity is as follows: (1) intensity variation due to the three-dimensional shape and material characteristics of the structure, (2) variation in intensity distribution due to the x- , And (3) fluctuation of the intensity distribution due to the unevenness of the surface of the structure.

So, I think about these three factors. First, the y-coordinate is fixed, and the cross-sectional shape when the pattern is cut in the xz plane and the one-dimensional signal intensity distribution corresponding to the shape are considered. In this case, the structure shown in the interrupted diagram (b) of Fig. 2 is a diagram showing the state of the macro structure shown in the left side of Fig. 3 (a) The concavo-convex structure of a micro-surface as shown schematically on the left side of FIG. 3 shows an image in which a macro pattern and a micro pattern are superimposed on the left side of the lower portion (c) of Fig.

As described above, the secondary electron emission amount increases as the electron beam enters the convex portion, and conversely, decreases as the electron beam enters the concave portion. Therefore, in the macroscopic structure of the entire pattern, since the upper surface corner portion of the pattern is convex, the signal intensity increases toward the upper surface corner portion (the right side of the top view (a) of FIG. 3). On the other hand, in the case of the micro structure, the signal intensity is increased when it is incident on the protrusions of the irregularities existing on the pattern surface (the right side of the interrupted view (b) of FIG. 3). Accordingly, the secondary electron signal intensity distribution correspondingly corresponds to a large peak distribution corresponding to the macro pattern structure (the right side of the top view (a) in Fig. 3), a fine signal variation corresponding to the surface rugged structure (The right side of the bottom view (c) of Fig. 3) overlapped with each other. Note, however, that the angle of incidence of the electron beam with respect to the unevenness changes on the side surface of the structure, so that the signal variation due to the unevenness also changes.

Here, the distribution of the secondary electron signal intensity in the x direction from an arbitrary position in the y direction is considered. Since the surface irregularities are thought to occur at random, the corresponding signal intensity distribution also has uncertainty and deviations as indicated by the solid line or dot chain line on the right side of the bottom view (c) of Fig. In a length measuring SEM, typically, the x-coordinate when the obtained secondary electron signal intensity distribution is cut at a certain threshold value is detected as edge coordinates.

By changing the threshold value, however, a plurality of edges corresponding to different height positions z of the structure are detected. The edge coordinates fluctuate with uncertainty, and the characteristic of the variation? Is a function of the threshold value. That is, it is a function of the different height positions z (or position x in the vertical direction) of the structure.

In the following first embodiment, a method of estimating the solid (cross-sectional) shape z (x) of the structure from the characteristics of the function will be described. In the second embodiment, a method of estimating a two-dimensional solid shape using a two-dimensional signal intensity distribution will be described. The details of the second embodiment will be described in the fourth embodiment which will be described later.

(First Embodiment)

Hereinafter, details of a method of estimating the solid (cross-sectional) shape of the structure according to the first embodiment of the present invention will be described. Hereinafter, three types of samples A, B and C having cross-sectional profiles shown in the upper part (a), the middle part (b) and the lower part (c) of FIG. 4 are analyzed.

First, an upper-side observation image of a structure having the three types of cross-sectional profiles is acquired by CD-SEM, and an edge portion of the structure is designated as an analysis region. In the target edge, the straight line portion of the design is selected, and the direction along the straight line is the y direction.

Next, the edge is detected by varying the threshold value T at the time of edge detection and an edge point sequence {xi (T)} (where i = 1, 2, ..., n) is obtained as a function of T, The deviation σ ² (T) = Σ [xi (T) - <x (T)>] ² from the average edge position <x (T)> of the point sequence was calculated. Next, plotting? (T) with respect to the threshold value T results in the upper diagram (a), the interim diagram (b) and the lower diagram (c) of FIG. As described above, the row of points shown at each stage corresponds to a structure having the same cross-sectional profile as that of Fig.

Generally, the x-direction distribution I (x) of the detected signal intensity has a peak in the vicinity of the edge of the structure. For example, when a so-called line pattern having a convex structure with a certain width W is observed, such as a resist line pattern, two intensity peaks are shown corresponding to the left and right edges of each pattern shown in Fig. That is, when one edge is observed, two edge detection points are obtained on both sides of the distribution peak of the detected signal intensity with respect to the same threshold value T. Therefore,? (T) is obtained separately for each of the inside and outside of the structure of the peak.

Generally, at the outer side (space side) of the structure, the x-coordinate of the point at which the intensity gradually increases from the outside and reaches a predetermined threshold value is defined as an edge position. On the other hand, at the inner side (line side) of the structure, the intensity is gradually increased from the inside, and the x coordinate of the point reaching the predetermined threshold value is taken as the edge position, and? (T) is obtained from these values. Further, when the pattern width W becomes extremely small, the peaks with respect to the left and right edges overlap, and one peak is observed with respect to one line pattern. In this case, the threshold value and the x coordinate correspond one by one.

Next, the average distribution <I> (x) of the distribution I (x) of the detected signal intensities in the y direction is obtained in the analysis region and each of the threshold values T is calculated as follows: To a specific X-coordinate of the X-axis. Thereby, the deviation? (T) as a function of the threshold value T is converted into the deviation? (X) as a function of x. At this time, the deviation? (T) with respect to x on the outside / inside of the structure of the peak corresponds to x on the outside / inside of each structure. When the deviation sigma (x) is plotted together with the detected signal intensity distribution < I > (x), the characteristic shown in Fig. 6 is obtained. The characteristic diagrams shown in the upper part (a), the intermediate part (b) and the lower part (c) of FIG. 6 all correspond to the structural part having the same section profile shown in FIG. Each of Figs. 6A to 6C shows a detection edge variation at an average edge position obtained with respect to the electron beam incident on the position x.

In the case of the sample A having a large film reduction and also in a forward taper shape (upper view (a) in Fig. 4), the deviation sigma (x) is large on the flat surface of the outer lower portion of the structure, And then gradually increases toward the center of the structure and takes the maximum at the vicinity of the flat surface of the upper portion of the structure.

On the other hand, in the sample B (the interrupted diagram (b) of FIG. 4) and the sample C (the lower diagram (c) of FIG. 4) having the sidewalls of a substantially vertical or slightly inverted tapered shape, (x) takes a minimum value and then abruptly increases to take a peak in the shape of a ground, gradually decreases gradually toward the center of the structure after it is once decreased, and reaches a maximum at the vicinity of the flat surface of the upper part of the structure do. The height of the peak is slightly larger than that of the sample C having the reverse tapered side wall.

It is considered that this difference in the distribution shape reflects the difference in the cross-sectional profile.

Next, the meaning of the shape of the deviation? (X) will be described.

First, the three-dimensional shape Z (x, y) of the actual structure is calculated in consideration of the average sectional profile <Z> (x) for convenience, corresponding to the average detected signal intensity distribution <I> ) Is expressed by the following equation.

[Equation 1]

That is, the actual three-dimensional shape Z (x, y) is obtained by shifting the average cross-sectional profile Z (x) in the x direction by? X (y) along the edge point position y, y). Here,? X (y) is determined so that the x-direction integral value of |? Z (x, y) |

Next, the three-dimensional distribution of the detected signal intensity actually observed corresponding to the above expression can be expressed as follows.

[Equation 2]

Here,? I (x, y) is a term generated by a deviation from the average of the sectional profile corresponding to? Z (x, y). ? Noise (x, y) is a random detection noise superimposed on an actually detected image, and its amplitude does not depend on the position / image. ? X (y), for example, |? I (x, y) | ² is set to be the minimum in the vicinity of the edge in the x direction.

In addition, the deviation? (X) from the average of the detected edge point sequences corresponding to the above expressions can be also divided into the components attributable to the respective terms.

[Equation 3]

Here, σ_measured is the measurement result. ? y is a component due to the deviation of the edge position of the structure when the cross section is optimized for an average cross-sectional profile every y. ? xz is a component due to a change in the cross-sectional shape of the structure. ? noise is a component due to an edge detection error due to detection (image) noise.

(Details of decomposition method of edge variation component)

Next, a method of decomposing the measured edge variation value? _Measured into the above-described four components will be described. Hereinafter, each component will be described.

The variation in the y direction of the edge position of the structure is generally considered to be the spatial frequency characteristic of the measurement error and the detection noise caused by the surface irregularities, while the spatial frequency characteristic generally depends on the 1 / f characteristic. Here,? _Y is considered to be a component having a 1 / f characteristic with respect to the spatial frequency f in the y-direction (called "true LER" for convenience).

A method for decomposing the deviation of the edge point detection position in the x direction measured along the edge into a component having a 1 / f characteristic and other components is described in Patent Document 1. Here, the decomposition of the LER component may be performed for each threshold value, or may be performed for any representative threshold value T.

The true LER is the variation in the y direction of the edge position itself of the structure. Therefore, it can be considered that measurement results obtained by changing the threshold value T are almost common. Because of this, the latter idea is reasonable.

[sigma] noise depends on the slope of the signal strength distribution. Fig. 7 shows the result of obtaining the variation of the edge point obtained with respect to the threshold of the average level by adding an intensity gradient in the one-dimensional (x) direction to the image obtained with respect to the smooth flat surface. As shown in Fig. 7, when the intensity gradient of the detection signal increases,? _Noise decreases. sigma_noise (x) can be calculated by obtaining sigma_noise with respect to the gradient at the position x of the average detected intensity distribution from Fig.

By removing the contribution of σ_y and σ_noise from the measurement result σ_measured, σ_xz can be obtained.

Here, σ_y and σ_noise are calculated for the samples A, B and C having the cross-sectional profiles shown in the upper diagram (a), the interim diagram (b) and the lower diagram (c) of FIG. 4 and the assumed result σ_measured , σ_y, σ_noise, and σ_xz are shown in the upper, middle, and lower diagrams (a), (b), and (c) of FIG. As can be seen from Fig. 8, when the cross-sectional profile is an inverted tapered shape, a remarkable peak appears in? _Xz.

Hereinafter, in order to consider the cause of remarkable peaks, the influence of the surface irregularities on the detected signal intensity distribution is classified into the following two types.

[Category 1]

In this classification, the electron beam is first considered to be incident on the forward taper surface. At this time, it is considered that the number of electrons detected outside the structure after scattering in the structure is affected by the concave-convex pattern in the vicinity of the incident position on the forward taper surface.

[Category 2]

In this classification, the electron beam is considered to be incident on a relatively flat surface on the upper portion of the structure away from the edge point. At this time, a part of the electrons scattered inside the structure is extracted from the side surface or reverse taper surface of the structure to the outside of the structure and is detected. At this time, the electrons exiting to the outside are influenced by the concave-convex pattern of the side surface. In this way, the absolute number of electrons that escape to the outside from the plane different from the plane of incidence is small. However, since the number of electrons detected by surface scattering is relatively small, it is considered that the influence of electrons influenced by the concavo-convex pattern on the side on the detection result on the forward taper surface is not negligible.

This influence depends on the area of the vertical or inverse tapered surface in the range of the length of the electron beam entry length inside the structure, and is larger as the height of the vertical or reverse tapered surface is larger. The influence also extends to the electron beam incident on the relatively large area from the edge of the structure toward the inner side.

The peak of sigma_xz shown in the upper part (a), the middle part (b) and the lower part (c) of FIG. 8 is the peak of the peak . From this, it is considered that the peak of? _Xz appearing on the upper flat surface of the structure is due to classification 2.

Then, σ_xz is expressed by (Equation 4) which is decomposed into the following two components.

(1) Influence of the concave-convex pattern on the forward tapered surface near the electron beam incidence point

     : σ_xz_near

(2) Influence of the concave-convex pattern on the substantially vertical or inverse taper surface relatively far from the electron beam incidence point

     : σ_xz_far

[Equation 4]

Hereinafter, the details of the components (1) and (2) will be described in detail.

First, the influence of the unevenness (the component (1)) on the scattering of the electron beam incident on the forward tapered surface having the concavo-convex pattern will be considered. It is assumed that sinusoidal irregularities exist at an inclined plane at a predetermined angle, and the beam is incident on the center of the concave or convex portion (point P in the top view (a) of Fig. 9). When the phase of the concave and convex is changed 180 degrees with respect to the incident position of the beam, the detection position fluctuates by one cycle in the ± direction of the phase. For example, when the beam is incident at the center of the concave portion, the maximum displacement occurs in the positive direction of the detection position, and when the beam is incident at the center of the convex portion, the maximum displacement in the negative direction of the phase Lt; / RTI &gt;

Hereinafter, for the sake of simplicity, as shown in the top view (a) of Fig. 9, the average sidewall surface of the structure having the sidewall shape having a constant inclination angle &amp;thetas; Is considered. In the case where the convex center of the surface irregularities of the representative period L exists on the point P, when the electron is incident toward the point P, the increase of the detected signal intensity becomes maximum and the shift amount of the edge detection position becomes maximum It is said. Conversely, when the center of the concave portion of the surface unevenness exists on the point P, it is assumed that the decrease amount of the detected signal intensity becomes the maximum, and the shift amount of the edge detection position becomes the maximum in the + direction.

At this time, the shift amount of the detection position of the edge due to the unevenness is determined by the x-coordinate position of the center of the convex portion, and the fluctuation amplitude? X is the x-direction distance between the centers of the convex portions in the two cases. Assuming that the concave-convex period on the sidewall surface is L (that is, the distance in the x direction is L / 2) and the detected position is determined in the interrupted view (b) of Fig. 9, , The concavity and convexity period L, and the angle (inclination angle) &amp;thetas; of the side wall sloping surface.

[Equation 5]

On the other hand, assuming that the detection position is determined in the bottom view (c) of Fig. 9, the fluctuation width? X of the detection position of the edge can be estimated by the following equation according to the height H and the angle (inclination angle)

[Equation 6]

In general, Equation 5 becomes dominant in a region where the inclination angle &amp;thetas; is comparatively small, and Equation 6 dominates in a region where the inclination angle &amp;thetas; is relatively large. In practice, it is considered that the inclination angle? Depends on both the period of the concavity and convexity of the side surface and the height. Therefore, it is also possible to estimate the variation width? X of the edge detection position by adding the formulas and then by the following equation.

[Equation 7]

The relationship between the variation width? X of the edge detection position and the inclination angle? Of the inclined surface can be obtained by using a characteristic that the intensity variation of the electron beam signal due to the surface irregularities depends on the incident angle of the electron beam to the surface.

For example, the detected signal intensity distribution when irradiating and scanning electron beams having various incidence angles to a plane (uneven view (a1) in Fig. 10) having concave and convex portions of appropriate period L or height H on the surface is calculated by simulation (Upper view (a2) in Fig. 10). At this time, the variation width? I of the detected intensity is a function of the incident angle? (FIG. 10, the interrupted view (b)).

On the other hand, the relationship between the intensity fluctuation width? I (?), The fluctuation width? X of the edge detection position, and the average signal intensity distribution I (x) can be expressed by the following equation as shown in the lower diagram (c) of FIG.

[Equation 8]

Equation (8), however, emits a value when the gradient becomes 0 (zero) at the peak of the intensity distribution of the detection signal. Therefore, it is necessary to note that Equation 8 can not be used when the gradient of the intensity distribution of the detection signal is 0 (zero). Actually, the amplitude? X of the edge shift amount is a statistic estimated from the measurement deviation of the edge detection position due to the unevenness of the side wall surface, and? _Xz_near corresponds to this. Since the concavo-convex pattern of the side wall surface is a random amount, a statistical representative value is also used for the period L and the height H. For example, for convenience, the inclination angle? Of the flat surface may be selected to be zero.

Next, the component (2) (the influence of the concavo-convex pattern at the surface position away from the incident position of the beam) will be described.

When the detected intensity when the electron beam is incident on the Q point in the upper diagram (a) of FIG. 11 is IQ with respect to the average sidewall surface, as shown by the curves A and B in the figure, We consider the case of fluctuation. At this time, as shown in Fig. (B) at the bottom of Fig. 11, the detected intensity is varied by ΔIA and ΔIB against the IQ detection strength, edge detection is shifted by Δx _A, Δx _B, respectively.

At this time, if the spread of the electron beam is about the same as the size of the sidewall surface, then the displacement amounts? X _A and? X _B appearing at the detection edge can be considered to be substantially constant without depending on the incident position (Q point).

Here, assuming that the incident position of the electron beam is substantially constant in the range of the scattering electrons in the structure from the edge position of the structure to the sidewall, the influence of the above-mentioned (2) is almost constant, do.

The height of the trapezoid depends on the height and angle of the sidewall, but the distribution of? _Xz_near after subtracting? _Xz_far from? _Xz on the structure side from the minimum value of? _Xz increases substantially linearly from 0 (zero) As shown in FIG. The results obtained by decomposing sigma_xz into the above two components for each of the samples corresponding to the upper diagram (a), the intermediate diagram (b) and the lower diagram (c) of FIG. 4 are shown in the upper part (B) and the bottom view (c).

(Method of estimating cross-sectional shape)

Next, a method for estimating the cross-sectional shape of the structure from the result of decomposing the deviation? Indicating uncertainty will be described.

The actual cross-sectional shape of the structure is considered to have a round shape, a curved shape, a gentle extension, or various shapes. Namely, in general, the sidewall angle? Is not constant with respect to the position in the height direction or x direction of the structure. By using the method according to the present embodiment, the cross-sectional shape of such a structure can be appropriately estimated.

Hereinafter, the section estimation method by the two approaches of the analytical shape estimation method and the model-based shape estimation method will be described.

(Analytical estimation method)

In the analytical estimation method, the local angle of the surface of the structure at the corresponding height of the structure is obtained by using Equation 5, Equation 6, Equation 7, Equation 8 or the like with respect to the position in the threshold or x direction, Direction to obtain the cross-sectional shape of the structure. That is, the cross-sectional shape is obtained by the following formula. However, in the following equation, the integral range is 0 (zero) to x.

[Equation 9]

Here, the distribution of? (X) estimated from? _Xz_near in Fig. 12 and the estimated cross-sectional shape Z (x) are shown in the upper diagram (a) And (c), respectively. The estimated cross-sectional shape (the shape shown by the solid line) is separately found to be in good agreement with the shape obtained from the cross-sectional observation (the top view (a), the intermediate view (b), and the bottom view (c) .

On the other hand, in the shape estimation method using the model base, the distribution of the detected signal intensities when various irregularities (for example, phase) of the surface are varied for various cross-sectional shapes and the deviation Uncertainty) is determined in advance as a function of the threshold or position in the x direction. Next, this deviation (uncertainty) is matched with the actual measurement result, and the cross-sectional shape is estimated for the nearest cross-sectional shape or for interpolation / extrapolation.

The conventional MBL method for estimating the cross-sectional shape by obtaining the matching between the library and the measurement result as a result of calculation of the intensity distribution of the detection signal itself has a problem that the measurement result is influenced by the detection system or the variation of the detection intensity, Are hardly affected by these. Of course, this method and the conventional MBL method may be used in combination. The maximum likelihood method or the like may be used for matching.

(Structural solid shape estimation processing procedure)

Fig. 14 shows a flowchart of a method for estimating the three-dimensional shape of the structure corresponding to the first embodiment. A series of processes to be described later is realized based on a program executed in the computer.

First, a two-dimensional image is acquired by an electron microscope, and an analysis area is designated (step 1401). Thereafter, the average signal intensity distribution I (x) in the designated analysis area is calculated (step 1402). The signal strength is normalized to the maximum intensity in the specified range. At this time, it is preferable to adjust the image so that the average edge direction in the range is the y direction.

Next, the threshold value is set at a predetermined interval from the designated minimum value to the maximum value (for example, from 5% to 100% every 5%) to detect a pattern edge for each of the threshold values, and LER (Steps S1403 to S1408). Here, LER is transformed into a function of x using the relation of T = I (x), and then decomposed into components. Further, the sidewall angle? Is obtained in accordance with equations 5, 6, 7 and 8 (steps S1409 and 1410), and the sidewall angle? Is integrated in the x direction to obtain a sectional shape (step S1411). In the description so far, it is assumed that the concavo-convex pattern of the surface of the structure is isotropic on the surface. Regarding the irregularities derived from the local variation in the solubility of the resist and the like, the above assumption is almost correct. On the other hand, a case in which this assumption does not hold is described in the fifth embodiment to be described later.

[Example 1]

In this embodiment, an embodiment in which the above-described estimation method is applied to a CD-SEM will be described.

(Device Configuration)

15 is a schematic diagram of the hardware configuration of the CD-SEM used in this embodiment. The CD-SEM of this embodiment mainly comprises a casing 1801 of a scanning electron microscope consisting of an electron optical column (SEM column) and a sample chamber, a control system 1811 of a scanning electron microscope and an information processing apparatus 1812 .

The information processing apparatus 1812 is connected to a data storage device 1813 for storing the obtained scanning electronic image and CAD data necessary for analysis. Of course, the data storage device 1813 may be stored in the information processing device 1812. [ Although not shown, the information processing apparatus 1812 includes an information input terminal for inputting information necessary for data processing by the operator of the CD-SEM to the apparatus and image display means for displaying the scanned electronic image to be acquired . Specific examples of the information input terminal include a keyboard, a mouse, and a GUI screen displayed on the image display means.

The electron optical column is constituted by an electron gun 1802, a convergent lens 1804, a deflector 1805, an objective lens 1806, a detector 1810 and the like. The sample chamber has a stage 1808 on which an observation wafer 1807 to be inspected is placed. Secondary electrons 1809 generated by the electron beam 1803 irradiated from the electron gun 1802 to the observation wafer 1807 are detected by the detector 1810 and converted into digital data by the control system 1811, And is transmitted to the information processing apparatus 1812 to generate image data to be used for analysis.

In this embodiment, pattern observation was performed using a scanning electron microscope provided in a CD-SEM to obtain image data of a subject to be inspected. The obtained image data was stored in the data storage device 1813, and after the observation was completed, the image data analysis was performed by operating the information input terminal to analyze the roughness index and the cross-sectional structure estimation. The analysis processing is executed by the information processing apparatus 1812.

 (Image Acquisition Step)

First, the control system 1811 (information processing apparatus 1812) averages the secondary electron signal intensities obtained by scanning the line pattern of the ArF resist 32 times from the upper left to the lower right of the visual field, A two-dimensional distribution image of the electronic signal intensity is obtained. If necessary, the information processing apparatus 1812 displays the acquired image on the monitor screen of the CD-SEM. Here, the number of pixels of the observation image is 1500 pixels in the vertical and horizontal directions, 1 nm (one nanometer) on one side of one pixel, and 1.5 micrometers (micrometers) in the vertical direction.

The information processing apparatus 1812 sets a rectangular inspection area of 1024 pixels in the vertical direction and 50 pixels in the horizontal direction in the area including the edge to be analyzed out of the two-dimensional distribution image of secondary electron signal intensity. In the data series extraction, the minimum value Tmin, the maximum value Tmax, the increment value? T, the sampling interval? Y in the y direction, the noise reduction parameter in the x direction, and the averaging parameter S in the y direction, Set the necessary information. It is also possible to set the number of detections, not the sampling interval in the y direction in data extraction. It is preferable that these inspection regions and conditions of data series extraction are set in advance through, for example, a GUI screen of a CD-SEM.

Next, the information processing apparatus 1812 executes a task of extracting data series of edge roughness in the area. That is, the information processing apparatus 1812 calculates the signal intensity distribution corresponding to the y coordinate at the sampling position from the pixel data in the inspection region in accordance with the set extraction start timing and the sampling interval, and also calculates the signal intensity distribution corresponding to the maximum value Tmax The x-coordinate data of the edge point is calculated from the signal intensity distribution according to the threshold value T set for each increment value? (T) = {[Delta] xi (T)} in the inspection region as a function of the threshold value T, and finally, = {? X1 (T),? X2 (T), ... }.

Specifically, the information processing apparatus 1812 sets the y coordinate corresponding to the lower side of the inspection area as the y coordinate of the data extraction starting point, sets 1 nm (nanometer) as the sampling interval in the y direction , Positions (x1 (T), y1 (T)) of 1024 points at intervals of 1 nm (nanometers) as edge points in the inspection region, ... (xi (T), yi (T)), ... (x1024 (T), y1024 (T)).

Next, the information processing apparatus 1812 approximates the extracted point by the following straight line, and obtains the values of? And? Which are fitting parameters.

[Equation 10]

Next, regarding the coordinates of the edge point with respect to all the threshold values T, the information processing apparatus 1812 obtains the shift amount? Xi of the edge point from the straight line according to the following Expression 11, and calculates the edge roughness series X ) = {DELTA xi} T.

[Equation 11]

(Image processing step)

Next, the information processing apparatus 1812 obtains the LER for each threshold value from the edge roughness series X (T) = {DELTA xi (T)}, calculates the obtained LER by the measurement deviation? Lt; RTI ID = 0.0 &gt; sigma_xz. &Lt; / RTI &gt; For example, the method described in Patent Document 1 can be used as a method of obtaining the measurement deviation σ_y by the true LER and the measurement deviation σ_xz by the surface irregularity from the edge roughness series. The following is shown below.

As described above, in the measured LER, the component (LER of true) in which the power spectral density is inversely proportional to the square of the spatial frequency f and the other high-frequency components (noise) are superimposed. When the averaged processing is performed on the measured LER, the latter component is reduced. Therefore, as the parameter value indicating the degree of averaging processing is increased, the power spectrum density distribution in the high frequency range becomes inversely proportional to the square of f. Concretely, when the one-dimensional signal intensity distribution in the x direction at different y coordinates is averaged S in the y direction, the intensity of the random noise is reduced to 1 / S by averaging. That is, as the averaging parameter S increases, the frequency dependency of the power spectrum density approaches 1 / f ² . At this time, the density of the power spectrum obtained becomes the true LER.

Here, the value of the line edge roughness index obtained from the data averaged with the averaging parameter S is defined as? M (S), and the S dependency is fitted to the following expression.

[Equation 12]

However,? M (1) is the line edge roughness measurement value obtained from the data before the averaging processing,? Y is the y direction extraction interval of the edge point, and A is the fitting parameter. S and the extraction interval Δy of the edge points preferably satisfy 2SΔy <1 / f ₀ [nm (nanometer)]. Here, f ₀ is, in the case of a typical resist pattern, at a bending point of the spectrum of 0.008 nm ^-1 Or more.

Here, σ ₀ = σ obtained for the smallest S among S, which best describes the experimental results, is set as the true LER. In this case, the measurement error component? B independent of the spatial frequency is obtained by the following equation.

[Equation 13]

The information processing apparatus 1812 obtains the measurement error sigma_noise from the measured LER and removes the measurement error sigma_noise from the measured LER to obtain sigma_xz (t) representing the measurement error (uncertainty) due to projection of the concavo-convex pattern on the sidewall surface onto the xy plane (T).

[Equation 14]

Here, the sidewall angle? At the height corresponding to the threshold value T of the structure is obtained by the following equation.

[Equation 15]

Here,? S is a representative value of the spatial period of the surface irregularities.

On the other hand, the average <I> (x) of the signal intensity distribution in the x direction was calculated with respect to the measurement range in the y direction. The edge coordinate x corresponding to the threshold value T can be obtained from x = < I > ^-1 (T). However, <I> ^-1 is the inverse of <I>.

By the way, the sectional shape z (= Z (x)) of the structure can be obtained by the following equation as described above.

[Equation 16]

However, the integral range is from T = Ia (0) to T = Ia (x).

The sectional shape obtained in this way was compared with the sectional shape obtained by cutting the pattern portion of the wafer and observing with SEM. Also, comparison with measurement results by AFM was also made, and it was confirmed that these were also in good agreement.

In the above description, σs and σ_noise are values having physical meaning and are independently measured and obtained. However, these quantities may be considered as fitting parameters. That is, σs and σ_noise may be used in which the cross-sectional shape observed by another method and the cross-sectional shape estimated by the present invention are in good agreement.

[Example 2]

In this embodiment, an example of a method capable of estimating a three-dimensional structure not only for a cross-sectional structure of a one-dimensional mask pattern but also for a two-dimensional mask pattern will be described. The configuration of the scanning electron microscope used in this embodiment is the same as that of the first embodiment, and a description thereof will be omitted.

The following two methods can be considered as a method of extending the method of the first embodiment into a two-dimensional mask pattern.

[First Method]

In this method, the relationship between the threshold value and the pattern height is obtained for the one-dimensional pattern by the method of the first embodiment, and this relationship is directly applied to the two-dimensional pattern.

[Second Method]

In this method, the method of the first embodiment is extended with respect to the two-dimensional edge information.

First, the first method will be described. The information processing apparatus 1812 applies the method of Embodiment 1 to the one-dimensional pattern portion of the obtained two-dimensional image (upper left view in Fig. 16) in the same manner as the method of Embodiment 1 to obtain the sectional shape z = Z ( x) is estimated (estimated).

Next, the information processing apparatus 1812 erases x from the two relations of z = Z (x) and the signal intensity distribution T = I (x) used for deriving the shape to obtain the height z And the threshold value T, z = Z '(T). When the edge is extracted by changing the threshold value T with respect to the SEM observation image of the two-dimensional mask pattern of the upper left diagram in Fig. 16, contour lines (contour lines) similar to those at the upper half of Fig. 16 are obtained.

Next, the information processing apparatus 1812 converts T in the upper middle diagram of Fig. 16 into z according to the relationship between the height z of the structure and the threshold value T, Get height information. Between T and z, the relationship shown in the lower part of Fig. 16 is established. The data giving this relationship is stored in advance in the data storage device 1813, for example.

The problem of this method is that there is not necessarily a guarantee that the relation z = Z (T) between the height z and the threshold value T of the structure obtained by the method of the first embodiment is satisfied in the structure of the two-dimensional mask pattern. Generally, the relationship between height and signal strength is not unconventional. However, for a pattern having a close pattern characteristic such as a line width, the present method can give comparatively good approximation.

Next, the second method will be described. The present method can be applied when the same design pattern exists in a plurality of different positions on the same mask. First, the information processing apparatus 1812 obtains a plurality of SEM images (signal intensity distribution) by observing a plurality of two-dimensional patterns on a wafer formed on a plurality of identical two-dimensional patterns on a mask under the same conditions by SEM observation .

For the plurality of SEM images, the information processing apparatus 1812 obtains an edge coordinate string at a certain threshold value T, and considers contour lines after each coordinate point. At this time, in order to secure the measurement accuracy, it is preferable to extract the edge by finding a signal intensity distribution along a direction substantially perpendicular to the contour line and applying a threshold value thereto.

Next, the information processing apparatus 1812 shifts each image in the parallel direction so that the sum of the distance deviations between the contour lines obtained for the plurality of images is minimized. The reference of the shift position may be set appropriately. The information processing apparatus 1812 calculates an average shape of the contour line of each image after the parallel shift and calculates the difference (distance) between the contour line and the average shape in the direction perpendicular to the tangent line of the contour line .

The information processing apparatus 1812 calculates the standard deviation of the distribution of the differences for each edge point and assumes a value corresponding to? B (T) in the first embodiment. Accordingly, as in the first embodiment, the inclination angle? Obtained by using equations 5, 6, 7, 8, etc. is defined as the sidewall angle? Of the height corresponding to the threshold value T at the position. The vertical structure along the above direction can be obtained by carrying out integration similar to the expression (9) along the direction in which the edge is extracted by changing the threshold value T. [

It is feared that the σb (T) obtained by the second method includes a component corresponding to the true LER. By performing the parallel shift, however, the edge shift due to the LER of a comparatively long period is canceled It is expected to be able. It is preferable that the short period component of the true LER and the measurement error calculation component are appropriately removed from the above-mentioned? B (T).

In the case of this embodiment, as in the case of the first embodiment,? S and? E are values having physical meaning and can be obtained by measuring them independently, or may be considered as fitting parameters. The inventors confirmed that the cross-sectional shape of the structure estimated by this method was in good agreement with the cross-sectional shape observed as an SEM image by cutting the same structure.

[Example 3]

In the present embodiment, the present invention is applied to a process for forming a resist pattern by photolithography used for manufacturing a semiconductor integrated circuit or the like, or for evaluating the quality of a three-dimensional structure of a semiconductor integrated circuit formed using the same, An example will be described.

In the present embodiment, a value obtained by integrating the absolute value of the deviation of the height of the estimated pattern and the height of the designed pattern on the entire surface of the evaluation region or a root mean square value of the evaluation value is used as the quality index of the formed pattern The amount corresponding to the area of the portion indicated by hatching in the right side view of the top view (a) of Fig. 17).

In addition, the cross sectional area Sp of the cross-sectional shape estimated by applying the method of the embodiment described above (for example, the lower part of the curve in the left side view of the top view (a) in Fig. 17) and the ideal design shape (Ratio) Sp / Si (in the case of the one-dimensional mask pattern) of the cross-sectional area Si of the rectangular shape in the upper view of FIG. 17 (left side view in FIG. The ratio Vp / Vi of the volume Vi of the ideal design shape (in the case of the two-dimensional mask pattern) may be used as the pattern quality index. However, in this case, it is preferable to treat the area or the volume of the pattern existing in the area where the design pattern should not originally exist as a negative value. Thus, even when a pattern exists in an area where a design pattern should not exist, it can be detected as quality deterioration.

As a result of optical simulation on a mask pattern obtained by performing optical proximity effect correction on the design pattern of the circuit pattern and the circuit design pattern, the ideal design form for the two-dimensional mask pattern includes various actual exposure results, And the maximum value of the volume estimated by applying the method described in the above.

A resist pattern for the same pattern on the mask for each of the exposure shots was observed by CD-SEM on wafers exposed by varying the focus and exposure amount conditions of the pattern forming exposure apparatus in a matrix form for each exposure shot on the wafer , The pattern dimension was measured for each shot from the observation image, the pattern cross-sectional shape was estimated by the method described in Example 1, and the pattern quality index was obtained.

As shown in the interim diagram (b) of FIG. 17, although the pattern dimension is monotonously decreased with respect to the exposure amount, the pattern dimension variation with respect to the focus setting value is small, and it is difficult to estimate the focus setting value from the pattern dimension variation. On the other hand, as shown in the lower diagram (c) of FIG. 17, the pattern quality index changes almost monotonously with respect to both the exposure amount and the focus, and was applicable as a monitor of the focus setting value. However, the pattern quality index is also changed with respect to the exposure amount. Therefore, it is preferable to first estimate the exposure amount from a change in the pattern dimension, and estimate the focus setting value using the pattern quality index with respect to the exposure amount.

As in the present embodiment, by using the estimated cross-sectional shape in the quality index of the formed pattern, it is possible to determine the optimum manufacturing conditions at a simple and high speed, and to detect the quality deterioration. Further, if the determination result or the like is fed back to the manufacturing process, quality deterioration of the formed pattern can be suppressed and the performance of various devices including the semiconductor device can be improved.

[Example 4]

In this embodiment, as another method of obtaining? X and? S, a method of performing frequency analysis on the two-dimensional signal intensity distribution in the xy direction will be described. That is, a method of estimating a two-dimensional solid shape according to the second embodiment will be described.

Here, assuming that isotropic surface irregularities are present along the surface on the entire surface of the side wall of the three-dimensional structure as shown in the upper part (a) of Fig. 18 and assuming that only the irregularity information is observed from above the three- do. At this time, it is considered that the pattern shown in the bottom view (b) of Fig. 18 is acquired as the unevenness information.

Here, when the period of the concavities and convexities in the x direction is Lx and the period in the y direction is Ly, the inclination angle? Is obtained by the following equation.

[Equation 17]

At this time, the two-dimensional intensity distribution of the secondary electron signal observed is firstly the macroscopic intensity distribution in the x direction, secondly the intensity variation due to the true LER in the y direction, and thirdly the local intensity variation due to the surface irregularities , And 3 variables. Therefore, by removing the first and second variation factors from the observation image, only the third surface irregularity information can be extracted. An example of a specific sequence will be described with reference to Fig.

First, the information processing apparatus 1812 acquires a SEM image, designates an analysis area, and acquires a two-dimensional intensity distribution within the area (step S1901). Next, the information processing apparatus 1812 applies the two-dimensional spatial frequency filtering to the acquired two-dimensional intensity distribution, and calculates a sum of the sum of the portion excluding the high-frequency component of the portion due to the first fluctuation component and the high- Frequency component of the second fluctuation component and the portion due to the third fluctuation factor (step S1902). The latter includes fluctuations due to high frequency components of the variation due to true LER and surface irregularities.

Generally, since the sidewall angle? Of the structure changes in the x direction, the x direction spatial period of the intensity distribution variation changes in the x direction. Therefore, it is preferable to evaluate the spatial period locally with respect to the x direction. As a general method for carrying out such an analysis, for example, there is a wavelet analysis. The information processing apparatus 1812 detects wavelet analysis in the x direction to detect a change in the spatial frequency characteristic along the x direction (step S1903).

An example of the x dependency of the spatial frequency characteristic (spatial frequency dependency of power spectral density PSD) obtained by this method is shown in Fig. When the peak is obtained, the frequency of the peak position is set as the representative spatial frequency, and when the peak is not clear, the spatial frequency characteristic is changed in the distribution area of the spatial frequency (for example, half width ) Is set as the representative spatial frequency, and the spatial period? X in the x direction is determined by the inverse number. Similarly, the information processing apparatus 1812 obtains? Y from the representative spatial frequency obtained by performing spatial frequency analysis in the y direction. Further, the information processing apparatus 1812 may perform a two-dimensional wavelet analysis in both x and y directions. In this process, if necessary, the information processing apparatus 1812 estimates the high-frequency component of the second variation component and removes it from the latter. In this way, the information processing apparatus 1812 obtains a two-dimensional distribution of? X and? Y.

The information processing apparatus 1812 obtains the two-dimensional distribution? (X, y) of the sidewall angle from the following equation (step S1904).

[Equation 18]

Further, the information processing apparatus 1812 estimates the two-dimensional solid shape by integrating the above equation at the position (x, y) (step S1905).

[Example 5]

In this embodiment, the case where the concavo-convex pattern formed on the surface of the structure is not necessarily isotropic will be described. In a case where the assumption that the concavo-convex pattern on the surface of the structure is isotropic is not established, for example, the shape of the side wall of the resist structure may be influenced by the standing wave due to the interference of exposure light in the resist film. In this case, the secondary electron signal intensity distribution is observed as a plurality of edges parallel to one edge of the structure, as shown in Fig. In this case, the sidewall inclination angle between the plural edges is expressed by the following equation, with the edge interval being? Lx.

[Expression 19]

Lambda is the wavelength of light used for exposure of the resist pattern, and n is the refractive index of the resist material with respect to the light of the wavelength. The standing wave appears when the reflection from the resist underlayer film is large at the time of resist exposure, but in this case, the resist dimension fluctuates sensitively according to the fluctuation of the resist film thickness. This is not preferable in an actual production process, and therefore, a sufficient antireflective measure is provided in order to suppress this. Therefore, in reality, the standing wave as described above affects less.

There is a case in which a resist pattern having isotropic surface irregularities is etched to transfer the etched structure as a case other than the one in which the irregularities on the surface irregularities are not necessarily established. The roughness of the surface of the resist pattern is transferred to the surface of the structure of the etched film by etching. According to the roughness analysis of the structure after etching transfer, roughness of the etched structure substantially reflects the roughness of the resist structure many.

In this case, as shown in Fig. 22, the period in the edge direction (y direction) of the unevenness does not change, but the dimension in the vertical direction changes. Assuming that the ratio of the etching rate Vr of the resist material to the etching rate Ve of the etched film is Retch = Ve / Vr and the etching is completely anisotropic, the longitudinal dimension of the concavities and convexities is approximately Retch times . In this case, assuming that the sidewall angle is relatively steep, L in Expression (5) may be multiplied by Retch.

Actually, the etching also proceeds in a direction (x direction) perpendicular to the edge direction (etching). In this case, the aspect ratio of the irregularities is not necessarily the same as Retch. Therefore, for example, it is preferable to optimize L in the expression (5) as a fitting parameter so that the cross-sectional shape observed by another method and the cross-sectional shape estimated by the above-described method are in good agreement.

The most general concept of the present invention is to estimate the three-dimensional shape of a structure from information of local variations in the signal intensity distribution of a two-dimensional image observed from the top surface of the structure. Therefore, the estimation algorithm is not limited to the above-described method.

(Another embodiment)

The present invention is not limited to the above-described embodiments, but includes various modifications. For example, the above-described embodiment has been described in detail with respect to some embodiments in order to facilitate understanding of the present invention, and it is not necessarily required to include all the configurations described above. It is also possible to replace some of the embodiments with the configurations of other embodiments, and it is also possible to add configurations of other embodiments to the configurations of some embodiments. It is also possible to add, delete or replace other configurations with respect to some of the configurations of the embodiments.

The above-described components, functions, processing units, processing means, and the like may be partially or entirely realized as hardware other than, for example, an integrated circuit. Further, the above-described configurations, functions, and the like may be realized by analyzing and executing a program that realizes the respective functions of the processor. That is, it may be realized as software. Information such as a program, a table, and a file that realize each function can be stored in a storage device such as a memory, a hard disk, a solid state drive (SSD), or a storage medium such as an IC card, an SD card, or a DVD.

Note that the control line and the information line indicate what is considered necessary for explanation, and not all the control lines and information lines necessary for the product. In practice, it can be considered that almost all configurations are connected to each other.

1801 Casing of scanning electron microscope
1802 electron gun
1803 electron beam
1804 compliant lens
1805 deflector
1806 objective lens
1807 observation wafer
1808 Stage
1809 Secondary electron
1810 detector
1811 Control system
1812 Information processing equipment
1813 Data storage device

Claims (10)

On the calculator,
Irradiating an upper surface of the substrate with a convergent energy beam from a direction substantially perpendicular to the main surface of the substrate having the three-dimensional structure formed on the upper surface thereof, scanning the upper surface of the substrate with a secondary energy ray A process of detecting and / or measuring the intensity of an energy ray reflected or scattered by the substrate and the structure and obtaining a top view of the structure,
A process of obtaining uncertainty information of scattering intensity due to the concave-convex shape of the surface of the structure from the irradiation position of the convergent energy line on the top surface observation and the measured intensity,
A step of obtaining an inclination angle? Of the surface of the structure based on the obtained uncertainty information,
A process of estimating the three-dimensional shape of the structure based on the obtained inclination angle?
Is performed. The method according to claim 1,
Wherein the convergent energy beam is a convergent electron beam, and the secondary energy beam is a secondary electron beam
And the pattern shape evaluation method. The method according to claim 1,
The process for obtaining the uncertainty information may include:
A plurality of edge point arrays corresponding to a plurality of levels having different intensities of the secondary energy rays are extracted and a deviation of each of the plurality of edge point arrays from the design coordinates is calculated to obtain a variation value of the edge point A sub-
A sub-process for obtaining an inclination angle of the surface of the structure at each position above the upper surface observation corresponding to the plurality of different levels based on the variation value
And the pattern shape evaluation method. The method of claim 3,
The sub-process for calculating the variation value of the edge point,
A step of obtaining a component that does not depend on the spatial frequency of the plurality of edge point sequences,
A step of removing a component whose power spectrum is inversely proportional to the square of the spatial frequency from the edge point sequence,
A step of removing the detected noise component
And the pattern shape evaluation method. The method according to claim 1,
The processing for obtaining the inclination angle?
As a function of the uncertainty information [sigma] x, an inclination angle &amp;thetas; with respect to the main surface of the substrate in the region is obtained within any predetermined region on the top surface observation
And the pattern shape evaluation method. The method according to claim 1,
The processing for obtaining the inclination angle?
Wherein a plurality of threshold values different from each other are set for the distribution of the intensity on the top surface observation and a tilt angle? Obtained by extracting a plurality of edge point rows with respect to each threshold value is set to a value at each edge position of the plurality of edge point rows The inclination angle?
The processing for estimating the three-
A three-dimensional shape is estimated by obtaining a tilt angle distribution at each point on the top surface observation and integrating the distribution of the tilt angle
And the pattern shape evaluation method. The method according to claim 1,
The process for obtaining the uncertainty information may include:
A sub-process for obtaining a tilt angle of the surface of the structure in the local region based on a difference in spatial frequency characteristics in a direction perpendicular and parallel to an edge of an intensity variation of a secondary energy line in a local region within the top- Containing
And the pattern shape evaluation method. The method of claim 3,
The edge point sequence is a curve
And the pattern shape evaluation method. On the calculator,
A three-dimensional structure including a semiconductor device is irradiated with a convergent energy beam from a direction substantially perpendicular to a main surface of a substrate formed on an upper surface thereof, and is scanned on an upper surface of the substrate. A secondary energy ray generated from the substrate and the structure A process of detecting and / or measuring the intensity of energy rays reflected or scattered by the substrate and the structure and obtaining an image of the top view of the structure,
A process of obtaining uncertainty information of the scattering intensity by the projected shape of the surface of the structure from the irradiation position of the convergent energy line on the top surface observation and the measured intensity,
A step of obtaining an inclination angle? Of the surface of the structure based on the obtained uncertainty information,
A process of estimating the three-dimensional shape of the structure based on the obtained inclination angle?
Processing for estimating the manufacturing conditions in the manufacturing process of the semiconductor device based on the estimated characteristics of the three-dimensional shape
Is carried out. Irradiating a convergent energy beam from a direction substantially perpendicular to a main surface of a substrate having a three-dimensional structure formed on an upper surface thereof, scanning the upper surface of the substrate with a second energy beam generated from the substrate and the structure, A data processing section for detecting and / or measuring the intensity of the energy line reflected or scattered by the light source and acquiring an image of the top surface of the structure,
A data processing unit for obtaining uncertainty information on the scattering intensity due to the concave-convex shape of the surface of the structure from the irradiation position of the convergent energy line on the top surface observation and the measured intensity,
A data processing unit for obtaining an inclination angle? Of the surface of the structure based on the obtained uncertainty information;
And estimates the three-dimensional shape of the structure based on the obtained inclination angle?
And the pattern shape evaluation device.

Images